The 10 most popular machine learning frameworks used by data scientists
Data science jobs are among the most coveted careers in America, taking the no. 1 spot on Glassdoor’s Best Jobs in America list for the past three years, and boasting high average salaries for those with the right skill set. These professionals report high job satisfaction as well, according to a recent report from Figure Eight: 89% of data scientists said they love their job, up from 67% in 2015.
Demand for data scientists remains high, the report found: 49% of the 240 data scientists surveyed said they get contacted at least once per week for a new job. Part of the reason for this is more companies are expanding their collection and use of data, and need a professional who can parse through it to drive business insights and apply it to new technologies like machine learning and artificial intelligence (AI).
Some 90% of data scientists say that some of their work informs AI and machine learning projects, the report found. Nearly 40% said that a majority of their work does so. However, that doesn’t mean the work is without its challenges: Some 55% of respondents said that the quality and quantity of training data is the biggest challenge in their work.
Data science and machine learning remain relatively young fields, so there is not yet overwhelming consensus on what languages, tools, and frameworks are best. While the machine learning community has largely begun using Python (61%), according to the report, there remains wide variability in the machine learning frameworks in use.
Here are the top 10 machine learning frameworks used by data scientists, according to the report:
- Pytorch & Torch
- AWS Deep Learning AMI
- Google Cloud ML Engine
It’s worth highlighting that many of these popular tools are open source—including Pandas, Numpy, Scikit-learn, Matplotlib, and Tensorflow—indicating that this community prefers open source, community-driven software, the report noted. Since many of these frameworks have been around for years, it’s likely that early adopters are now very familiar with them, and that it will take time, effort, and quality of performance for others to unseat them from the top slots.